SQL Prompt Generation Using Query Logs and Data Profiling

Resolve Bottlenecks,
Find Innovative SolutionsGenerate Solutions

Solution Overview

Problem

Existing methods for generating text-to-structured query language (SQL) queries using large language models (LLMs) face challenges in handling complex business databases with poorly documented tables and unclear data constraints, requiring costly human expertise for prompt engineering that does not scale well.

Innovation Solution

Automatically infer metadata from query logs and perform database profiling to generate prompts for LLMs, using query features and statistical analysis to improve the generation of SQL queries for complex databases.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If human expertise is used for prompt engineering to handle complex business databases, then query generation accuracy is improved, but cost and scalability deteriorate

Engineering Contradiction:

Improvequery generation accuracyVSAvoidscalability

Core Design Contradiction:

Measurement precisionVSProductivity

Solution Approach 1:

The system performs self-service by automatically inferring metadata from query logs and performing database profiling without human intervention. The processing system autonomously generates prompts for language models by analyzing historical query patterns and database structures, eliminating the need for costly human prompt engineering while maintaining scalability.

Inventive Principle:

Principle #25Self-service

Solution Approach 2:

The patent introduces an intermediary processing system that acts as a mediator between the complex business database and the language model. This intermediary automatically infers metadata, constructs prompts, and translates user intent into SQL queries, bridging the gap without requiring direct human expertise in prompt engineering.

Inventive Principle:

Principle #24Intermediary (Mediator)

2Measurement precision

If manual prompt engineering is used for complex databases, then query accuracy is improved, but time consumption and cost increase

Engineering Contradiction:

Improvequery accuracyVSAvoidtime consumption

Core Design Contradiction:

Measurement precisionVSLoss of time

Solution Approach 1:

The system performs preliminary actions by automatically inferring metadata and constructing prompts before the language model generates SQL queries. The processing system pre-processes database schemas, query logs, and data profiles to create optimized prompts, eliminating the need for time-consuming manual prompt engineering while maintaining high query accuracy.

Inventive Principle:

Principle #10Preliminary action

3Ease of manufacture

If existing LLM methods are used without metadata inference, then implementation simplicity is maintained, but query generation effectiveness deteriorates for complex databases

Engineering Contradiction:

Improveimplementation simplicityVSAvoidquery generation effectiveness

Core Design Contradiction:

Ease of manufactureVSReliability

Solution Approach 1:

The patent segments the query generation process into distinct automated components: metadata inference from query logs, database profiling, prompt construction, and language model execution. Each segment handles a specific task automatically, maintaining implementation simplicity while significantly improving query generation effectiveness for complex business databases.

Inventive Principle:

Principle #1Segmentation

Data Source

PatentUS20260050591A1Text-to-structured query language query generation using query logs and data profiling

Publication Date: 2026.02.19 AT&T INTELLECTUAL PROPERTY I L P

AI summary

In one example, a method performed by a processing system including at least one processor includes automatically inferring metadata about a structured query language database, constructing, based on the metadata, a prompt for generating a new structured query language query to be executed against the structured query language database, and providing the prompt to a language model for use in generating the new structured query language query.